data, analytics, & ai: how trust delivers value · analytics practices to incorporate ai-based...
TRANSCRIPT
ON BEHALF OF:
Findings From the Annual Data & Analytics Global Executive Study
Data, Analytics, & AI: How Trust Delivers Value
C U S T O M R E S E A R C H R E P O R T
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
Table of ContentsExecutive Summary .......................................................................................................................................................................................1
Section I: Building Trust in Data: How Analytics Leaders Get ‘the Right Stuff’ ...................................................................................2
• Trust Advances Analytics Maturity ......................................................................................................................................................2• Why Closing the Trust Gap Matters .................................................................................................................................................... 3
• Grade Your Data ................................................................................................................................................................................... 4
• The Human Factor: Partner With Domain Experts ............................................................................................................................ 5
• AI Built on a Bedrock of Data Governance ........................................................................................................................................ 5 • Commit to Treating Data as an Asset ................................................................................................................................................. 6
Section II: Success With Customer Data Depends on Keeping Customers’ Trust .........................................................................10
• The Pivotal Roles of Data Security and Privacy ................................................................................................................................10
• The Opportunity to Build Customer Trust Based on Data ................................................................................................................11
• Trust Is Fragile — Handle With Care ...................................................................................................................................................11
Section III: Building Trust in Innovation by Creating a Culture of Inquiry and Experimentation ......................................................15
• Driving Data Literacy Through the Workforce ..................................................................................................................................16
• Fostering Collaboration Drives Culture Change ...............................................................................................................................16
• Analytics Expertise: Centralize vs. Decentralize .............................................................................................................................. 17
• Communication and Education Encourage an Analytics Mindset ..................................................................................................18
Leaders’ Best Practices
• Health Care – Cleveland Clinic’s Centralized Data Store Helps Build Trust in Analytics ..................................................................... 8 • Manufacturing – Caterpillar Tailors Analytics Strategies to Business Unit Needs .......................................................................... 9
• Government – DataSF Teaches the Art of Asking Analytical Questions ........................................................................................13
• Financial Services – Cross-functional Teamwork Improves Predictive Models at Barclays US ...................................................14
About the Research ....................................................................................................................................................................................20
Acknowledgments ......................................................................................................................................................................................20
Sponsor’s Viewpoint ....................................................................................................................................................................................21
MIT SMR Connections develops content in collaboration with our sponsors. It operates independently of the MIT Sloan Management Review editorial group.
Copyright © Massachusetts Institute of Technology, 2019. All rights reserved.
1
Deriving more value from analytics and emerging technologies
like artificial intelligence starts with trust, simply because data
collected for analytics must be trusted. Much like the need for
sterility in clinical laboratories or a clear chain of evidentiary
custody in law enforcement, customers and partners that share
data must trust that it’s safeguarded and used appropriately from
collection through storage and to how it’s applied. Conversely,
once insights emerge from applying analytics to the data, indi-
viduals throughout the organization must understand the care
given to data management so that they trust those insights —
and use them — to make decisions and ask new questions.
Our global survey of more than 2,400 business leaders and man-
agers provides insight into organizations’ activities in each of
these key areas and identifies where recognized best practices
are becoming more mainstream and where they may still be
exceptional. It found that respondents who have advanced their
analytics practices to incorporate AI-based technologies such
as machine learning and natural language processing work in
organizations that do the most to foster data quality, safeguard
data assets, and develop cultures of data literacy and innovation.
Overall, the survey found a persistent gap: While the majority
of respondents report increased access to data, only a minority
say they have the “right” data to make decisions. Notably, those
who report a high level of trust in data for analytics are also
more likely to show leadership in foundational activities that
ensure that the data is high quality and leads to useful insights.
The survey, augmented by interviews with leading practitioners
and experts in the field, gauges the data analytics practices
of organizations around the world and how respondents view
the effectiveness of those practices. Three major conclusions
emerge among the chief findings:
1. Better data governance is needed.
A minority of respondents have formal activities in data quality
assurance, which points to the need for a greater commitment
to data governance in support of advanced analytics. The prac-
titioners interviewed provide examples of approaches to data
architecture, governance, and quality that can build trust in
data for analytics.
2. Data privacy emerges as an opportunity.
Data security is the strongest focus among survey respondents,
but there are opportunities to increase the maturity of security
practices by applying analytics and AI in this arena. Data
privacy initiatives are not quite as strong. Interviews highlight
opportunities to approach privacy initiatives and GDPR as a
way to build trust with customers, rather than simply treating
them as compliance-driven mandates.
3. Fostering an analytics culture improves innovation.
Leadership and management practices that support a culture
of analytics-driven innovation are relatively strong, but the
research shows many organizations have an opportunity to do
more to spread the necessary skills and mindset throughout the
workforce. A minority of respondents are engaged in activities
that develop workforce data literacy; meanwhile, finding a
workforce with the right skills was cited as one of the most sig-
nificant challenges to innovating. Interviews with practitioners
show how analytics leaders can foster a culture of innovation
by educating, communicating, and collaborating with partners
in lines of business.
Organizational choices — such as centralizing the analytics
function and having a chief data officer or chief analytics
officer role — may also help advance analytics maturity. Those
who have a CDO or CAO are more likely to report that they
have the data needed for decision-making, as are those who
work in organizations where analytics are centralized.
For leaders of organizations still striving to achieve analytics
maturity, the testimony from practitioners — who range
from data leaders at multinational corporations to the head
of a small municipal government team — is particularly
useful. Their chief lesson: Communication and collabora-
tion between analytics and business experts lead to mutual
understanding and measurable benefits. In other words,
trust delivers value.l
Executive Summary
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
2
Building Trust in Data: How Analytics Leaders Get ‘the Right Stuff’
When enterprise information leaders at the Cleveland
Clinic set the goal of advancing the organization’s
data analytics maturity about four years ago, they
had already established strong programs for decision support
and business intelligence. They knew that going beyond dash-
boards and reports to give clinicians and managers predictive
— and prescriptive — insights powered by artificial intelligence
and machine learning would demand more than just supplying
the technology.
The leading academic medical center recognized that to create
a culture where people understand and use advanced analytics
to make better decisions, it needed to focus on its information
and data: “Ensure that it’s available, that it’s valid, that it’s
governed appropriately, that we have the right processes
around accessing it, using it, sharing it, protecting it,” says
Chris Donovan, executive director of enterprise information
management and analytics.
Trust Advances Analytics Maturity
Like other leading organizations pursuing advanced analytics
capabilities, the Cleveland Clinic has placed a priority on
building trust — trust in the data that’s collected and stored
and trust in the analytic insights it generates. And it has seen
how building that trust can reinforce a culture that trusts and
embraces data-driven decision-making.
Findings from our recent survey, which focused on the data
and analytics practices of more than 2,400 business leaders
and managers, underscores the importance of such priorities.
We found a strong correlation between those who report
using the most advanced analytics techniques and those whose
organizations actively foster data quality, safeguard data
assets, and build a data-driven culture.
These organizations focus on data quality and management.
They set up measures for governing its proper use and security,
and by following these practices, they achieve results. In the case
of the Cleveland Clinic, that means more researchers trust the
data they are accessing from a centralized data lab instead of
copying data to work on in their own, siloed system. This leads to
more consistent data — and more precise results, Donovan says.
However, such benefits, bred from best analytics practices, are
still not widespread: Our research shows that most organiza-
tions are still developing their analytics capabilities. Just 15%
of survey respondents report that they use advanced analytics
to inform management decisions. Fewer than one in 10 are
working with automated analytics, and only 7% apply machine
learning and artificial intelligence in decision-making or
production workflows. Far more common, respondents rely
on business intelligence tools and employee dashboards to
support decision-making.
At the same time, we observed what could be called a “utility
gap”: While 76% of respondents report they have increased
access to data they judge useful — which is not surprising, given
the proliferation of data that accompanies the digitalization of
business — that access does not equal empowerment. A much
smaller number say they are able to leverage that data: Only
43% feel they frequently have the right data needed to make
decisions. This utility gap is a persistent trend, with a similar
gap found in the MIT Sloan Management Review survey in 20171
(see Figure 1).
While the majority of survey respondents reported increased access to data in surveys conducted in 2017 and 2018, those who believe they have the data they need for decision-making remain in the minority.
Figure 1: A ‘Utility Gap’ Persists
1 S. Ransbotham and D. Kiron, “Using Analytics to Improve Customer Engagement,” MIT Sloan Management Review, Jan. 30, 2018.
78% 76%
44% 43%
2017 2018
Percentage of respondents reporting somewhat or significantly improved access to useful data over the past year
Percentage of respondents reporting frequently or always having the right data to informbusiness decisions
Percentage of respondents reporting somewhat or significantly improved access to useful data over the past year
Percentage of respondents reporting frequently or always having the right data to informbusiness decisions
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
3
Why Closing the Trust Gap Matters
While there can be a range of reasons why people may not feel
they have the “right” data to handle, our survey probed and
found evidence for one: a trust gap. Only a very small minority
of respondents say they “always” trust data judged by qualities
of relevance, completeness, timeliness, and accuracy; slightly
more than half say they judge data trustworthy by those qualities
at least “often” (see Figure 2). This finding suggests a significant
opportunity to shore up data quality to build confidence in data
for analytics, in particular when it comes to increasing the
likelihood that data is complete — the aspect trusted least often.
There is ample opportunity for organizations to do more:
Only 21% of survey respondents report formal approaches
to data quality, which we defined as routinely monitoring,
managing, and improving data quality as part of a formal data
governance effort (see Figure 3). The largest group is reactive
to quality issues — a practice that Jeanne Ross, principal
research scientist at the MIT Center for Information Systems
Research, advises against.
“The worst place to fix the data is when it’s already been
collected,” says Ross. Data quality efforts should focus on the
business process that takes in data, whether that is from cus-
tomers or a part of the business.
Ross acknowledges such pragmatism takes commitment. While
it’s straightforward to say, “Fix your processes so that the data
collection is very reliable and the quality issues are pretty min-
imal,” meeting that goal is a challenge. That’s because it takes
ongoing discipline to refine data collection processes, testing
data quality regularly along the way.
But in her research, Ross has found the effort pays off. “Here’s
the interesting thing about analytics: Your unique opportunity
is going to be on your own data,” she says. The principle applies
whether a company is using its own internal data or augment-
ing that data with third-party sources. “If you have data, and
you supplement that data and you do that in ways that other
Few survey respondents are always confident in the quality of their analytics data, although a majority of respondents often trust that it’s accurate, up to date, and relevant. Trust in completeness of data is lowest, but trust in accuracy is most frequent.
Percentages may not equal 100 due to rounding.
Informal: Individuals who produce or use data reactively correct for accuracy, consistency, timeliness, and completeness
Data stewardship: Someone is responsible for proactively identifying and correcting causes of data quality problems
7%21%
30% 42%
Formal: Data quality is routinely monitored, managed, and improved as part of a formal data governance e�ort
No data quality e�orts
Just one in five organizations takes a formal approach to data quality, while 30% report at least proactive efforts. The plurality of respondents still tackle the issue informally.
Figure 3: Data Quality Efforts Show Room for Improvement
Figure 2: Data Accuracy Is Most Trusted Quality
How often do you trust that analytics data is:
40%
43%
Always
Relevant Complete Up to date Accurate
Often Sometimes Rarely Never
6%
1%
6%
28%
42%
21%
3%
12%
44%
34%
9%
1%
9%
47%
37%
6%
1%
11%
Always Often Sometimes Rarely Never
Informal: Individuals who produce or use data reactively correct for accuracy, consistency, timeliness, and completeness Data stewardship: Someone is responsible for proactively identifying and correcting causes of data quality problems
Formal: Data quality is routinely monitored, managed, and improved as part of a formal data governance effort
No data quality efforts
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
companies can’t, once you understand the data, once you get
insights, you’re able to do things other companies couldn’t.”
Grade Your Data
Our survey respondents recognize that data quality begins at
home: Internal data is most often subject to verification, fol-
lowed by customer data (see Figure 4). While that may lead to
the high degree of trust in internal data, customer data — key
to many advanced analytics applications — still lags in fourth
place when it comes to the degree of trust placed in it.
Practitioners and experts interviewed for this report note that
as more companies drive to build advanced analytics capa-
bilities and the volume of available data continues to expand,
organizations are looking for ways to include new informa-
tion sources in analytics systems. To do so effectively requires
building ways to mitigate the risk of bad data getting baked
into processes and driving faulty conclusions.
“The desire to do advanced analytics is driving this appetite for
more data, which makes people go out and pull data from other
places,” says David Loshin, principal consultant at Knowledge
Integrity. “Without instituting information governance prac-
tices, they’re at risk of not achieving the goals that they set out
to achieve.”
Trustworthy analytics requires setting up policies that assert
the standard for confidence levels in the data an organization
will use, determining the provenance of the data, and estab-
lishing acceptable data use, Loshin says. Statistical analysis and
technology tools can help identify problems and clean up data
sets (or lead to decisions not to use them).
Toronto-based Sun Life Financial generally begins with the
premise that no data is trustworthy, regardless of source, be-
cause all data has quality issues. Still, how data will be used also
factors into how trustworthy it needs to be, says Eric Monteiro,
senior vice president of client solutions at the global financial
services company. “We do hold different bars for different
types of use cases. For example, for general client segmenta-
tion or even business cases where we are sizing an opportunity
and making a strategic decision, the bar for quality is lower
because in general, the errors will even out up and down, and
you’re OK in the end,” he says. However, if data is used to drive
14%
42%
Regularly
Data from sensors/IoT
Sometimes Never
44%
62%
5%
33% 32%
51%
18%
39% 39%
21%
Internally generated
data
Publicly available
data
Regulators’data
34%
49%
18%
Competitors’data
42%48%
11%
Vendor-provided
50%
42%
9%
Customer-provided
4%
39%
Trusted
Data from sensors/IoT
Somewhat trusted Not trusted
57%
63%
2%
35%
23%
66%
11%
55%
40%
5%
Internally generated
data
Publicly available
data
Regulators’data
12%
67%
21%
Competitors’data
29%
64%
7%
Vendor-provided
37%
58%
5%
Customer-provided
Figure 4: Verify and Trust
How often do you verify: How much do you trust insights based on:
Most attention is paid to verifying internal and customer data. Internal data is also the most trusted source, while that provided by customers lags in fourth place.
Percentages may not equal 100 due to rounding.
Regularly Sometimes Never Trusted Somewhat trusted Not trusted
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
4
individual communications with clients — sending an email
based on the assumption of a certain life event, for example —
there must be less tolerance for error.
“If we go to you and say, ‘Congratulations on retirement,’ and
it turns out even though you’re 64, you are not at all thinking
about retirement, that’s pretty embarrassing and a pretty bad
client experience,” Monteiro points out.
The Human Factor: Partner With Domain Experts
Understanding what data can be trusted may be easier when
analytics practitioners work closely with domain experts in
their organizations.
Joy Bonaguro, who oversaw a team of five as chief data officer
for the city and county of San Francisco (she has left the orga-
nization since being interviewed for this report), says ongoing
consultations with in-house clients make or break a project.
That’s precisely because project teams are often using data
that was not collected with modeling in mind, she says.
“We definitely have to do clean-up, but the way we manage
quality is we probably spend an unusual amount of time com-
pared to your average data science team on doing what we call
exploratory analysis briefings,” Bonaguro says. “That’s where
we’re really confirming with a client, ‘Are we looking at your data
correctly? Where are there data quality issues? How should we
treat those data quality issues during the modeling phase?’”
Teaming data scientists with domain experts and data experts
— who understand data sources and how they can be auto-
mated — should be a best practice in every analytics operation,
says Dean Abbott, co-founder and chief data scientist at
SmarterHQ, an analytics firm that works with marketers.
For example, if an online retailer is seeking to understand data
showing a spike in abandoned online shopping carts, people
with different expertise and perspective may note different
patterns and identify different resolution paths that may have
been missed without that collaboration. A retail expert would
expect certain shopping behavior patterns and check analyt-
ics results against assumptions and identify questions to ask
that are not part of the initial probe. On the other hand, a
data infrastructure expert would be able to spot and examine
anomalies arising from technical issues and might know which
adjustments to the model would help either zero in on those
anomalies or provide the insight to confidently disregard them.
AI Built on a Bedrock of Data Governance
Those seeking to shore up practices that enable a more robust
capability in analytics and AI might look to construction and
mining equipment giant Caterpillar for inspiration.
Caterpillar uses AI to help marketers predict when customers
are ready to purchase and signal when customers are likely to
stop buying. Product development engineers use AI to simulate
product performance before building prototypes and
Teaming data scientists with domain experts and data experts — who understand data sources and how they can be automated — should be a best practice in every analytics operation. DEAN ABBOTT, SMARTERHQ
5
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
separately simulate how an equipment operator will use a
machine in the field so they can anticipate customer needs.
Natural language processing automates the analysis of legal
documents, checking the validity of agreements and pinpoint-
ing issues that need attention; it is also used to analyze tens of
thousands of customer and dealer comments to predict future
quality and warranty issues.
“Thanks to AI, things that would have taken a person two to three
weeks to do manually we can do in 10 minutes,” says Morgan
Vawter, chief analytics director at Caterpillar.
Behind all of these capabilities lies a strong data management
foundation, Vawter says. Data governance ensures that the
right data gets used so that results are trustworthy. Ensuring
that the data is high quality and its provenance is clear lets
analytics professionals start with good building blocks for
their models. “We need to make sure we have a common
truth in the data, including the underlying data, how it’s been
processed, how it’s been analyzed, how diverse it is,” she
says. “And then, ultimately, we can take that and build more
accurate, scalable analytics.”
At the Workplace Safety and Insurance Board (WSIB) Ontario,
charting the government agency’s course to AI and advanced
analytics begins with data architecture. Like many in her
position, Christina Hoy, vice president of corporate business
information and analytics, finds that challenge made more
complex by many legacy systems and data.
“We have a lot of data that’s in disparate systems. Up to now, we
have not had a formal strategy. We want to approach this in a
thoughtful manner,” Hoy says. She is working to create “a better
data strategy and an architecture that will actually let us use the
information the way it should be to make better decisions.”
WSIB is strong at diagnostic and descriptive analytics, Hoy says,
but the agency sees the opportunity to do more to op-
erationalize predictive analytics. It has also convened an
exploratory AI working group. Before it can take advantage of
such advanced technologies, Hoy says, “we need to make sure
our data is well-structured. You can’t do it if the foundation is
held together with duct tape and staples.”
Hoy invests time in communicating and educating to gain sup-
port and budget for this critical effort from within the agency
and from management. “If they can’t see what I’m trying to
achieve, I won’t be successful,” she says. “They have to be able
to visualize as well as I can what I’m trying to achieve. I think
that’s my biggest objective — to communicate that vision.”
Hoy sometimes uses an iceberg analogy, likening the 10% float-
ing above the water to the analytics results that leaders need,
while “the 90% of the iceberg that floats under the water is my
data architecture.” And she gets results: “They understand we
actually have to have a data strategy to enable analytics and
decision-making.”
Commit to Treating Data as an Asset
While many corporate leaders agree that data is an import-
ant asset, those who back up that view with committed
organizational resources may be faster to gain advantage from
AI and advanced analytics. At Barclays US, a set of management
decisions to treat data as an asset underlies the success of such
efforts, according to Vishal Morde, vice president of data sci-
ence and advanced analytics. For example, his business unit’s
chief data officer reports directly to the CEO. And the bank es-
tablished a data management council to catalog all data assets,
each asset’s owner, and policies for who gets to use the data.
This governance structure creates internal understanding
about the importance of sound data management and en-
sures trust in the company’s data resources, Morde says. The
“Thanks to AI, things that would have taken a person two to three weeks to do manually we can do in 10 minutes.” MORGAN VAWTER, CATERPILLAR
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
6
approach is crucial for the consumer bank’s Data Science
Center of Excellence, where the focus is on reaching customers
with compelling offers while maintaining a business relation-
ship built on trust.
Morde’s team has used advanced techniques such as machine
learning, deep learning, natural language processing algo-
rithms, topic modeling, and sentiment analysis to examine
customer complaints. After finding patterns in this text-based
data, the bank changed some of its policies.
“The natural language processing allowed us to uncover those
deep, hidden insights. We were able to think about it very com-
prehensively from the customer’s point of view and actually
made some tangible impact on the customer experience,” Mor-
de says. After these changes, complaint rates dipped to their
lowest point in four years. And in the 2018 J.D. Power Credit
Card Satisfaction Study, Barclays US moved from seventh posi-
tion to third position.
As these leading practitioners acknowledge, a focus on data
quality requires committing resources and budget. Again, our
research found a fairly modest minority in the vanguard: Just
15% reported significant budget increases for data quality in
the past year (see Figure 5). While it’s good news that 40% still
saw some increase in budget, it’s likely that the remainder,
whose budgets stayed flat or decreased, may find it a bit more
challenging to advance their analytics maturity.
Better news may be that a full third of survey respondents now
work in organizations that employ a chief data officer or chief
analytics officer. We found that those respondents are signifi-
cantly more likely to also report being on the right side of the
utility gap — meaning they have the right data to inform their
business decisions.l
Data quality needs funding to back up the commitment: Only a relatively small per-centage of companies gave these efforts a markedly higher priority in budgets over the past year.
Figure 5: Putting Their Money Where Their Data Is
40%
15%
38%
5%
2%
Significantlyincreased
Somewhatincreased
Nochange
Somewhatdecreased
Significantlydecreased
7
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
8
Cleveland Clinic’s Centralized Data Store Helps Build Trust in Analytics
I N D U S T R Y S N A P S H O T
The leaders of Cleveland Clinic’s 4-year-old enterprise analytics initiative are focused on building trust in the data the organization makes available to support decisions. Creating a central platform is one strategy to advance those efforts.
Before the effort began, Cleveland Clinic had a very decentralized approach, with teams building their own data stores and developing their own analytics projects with inconsistent results.
A centralized analytics platform was also about establishing one set of data for the organization, says Chris Donovan, executive director of enterprise information management and analytics. “If we make that platform compelling enough in terms of performance, and really partner with them and educate them in terms of the data that we have available, we eliminate that need to copy data all over the place and people begin to trust that central data store,” Donovan says.
The approach is working. Teams from around the clinic are moving some of their work into the platform, which is designed to enable them to have administrative rights to the data in a dedicated area of what Donovan calls a data lab. A team needs approval to create its space in the lab but not to do analytics work there.
There are some big benefits. First, teams that use the centralized platform are no longer asking Donovan’s group for copies of clinic data sets to experiment with. Second, the platform improves the quality of the data because people accessing it don’t have to worry whether they are getting the most up-to-date version of the data — that’s known. And third, the platform enhances the culture for analytics.
“Those very tangible changes in behavior indicate to me that we’re building that trust,” Donovan says.
Because people are using the central data store, Donovan’s team now has insight into updates and modifications that are made to the data. “We can have a conversation about, ‘Is this really the data that’s wrong? Is this an interpretation issue?’” he says. That kind of awareness can help clear up confusion about differing uses of terminology and lead to agreement about data definitions. By winning trust in a central data store, the clinic enabled a feedback loop that can lead to another benefit: better data management.
HE
ALT
H C
AR
E
Chris Donovan, executive director of
enterprise information management and analytics,
Cleveland Clinic
Those very tangible changes in behavior indicate to me that we’re building that trust.”“
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
9
“We have to make sure that we’re not skipping key parts of the journey,” Vawter explains. “You can’t just skip descriptive analytics, building dashboards, understanding data, and jump right to predictive and prescriptive analytics. We want to make sure that we’re helping them to understand their data at the foundation and then advance them up the maturity curve.”
With an overarching approach that provides a global view of all the company’s efforts, Caterpillar can take successful projects and apply them throughout the company. Previously, analytics excellence may have sprung up in pockets with efforts such as supply chain analytics or marketing analytics, but those tools may not have then been applied through the business unit. “Our analytics road maps have really showed the power of, ‘Hey, if I’m doing analytics here, it impacts this and it impacts the entire value chain.’ And so we’ve seen a lot of scaled analytics as a result of having those strategies,” Vawter says.
MA
NU
FAC
TU
RIN
G
Morgan Vawter, chief analytics director,
Caterpillar
I N D U S T R Y S N A P S H O T
Caterpillar Tailors Analytics Strategies to Business Unit Needs
To build a culture of data-driven decision-making and innovation, Caterpillar’s analytics group works closely with business unit leaders throughout the $45 billion company. This helps ensure that analytics supports business needs and wins executive support.
Morgan Vawter, chief analytics director at the company, says her team sits down with the leaders of units to tailor strategies to fit the needs of each. “It’s making sure that, because we have a very diverse business, we don’t just have an analytics strategy for the company — we have analytics strategies that enable the business unit strategies,” Vawter says. “Also, that’s useful because it gets that executive buy-in. They become advocates for it, and that cascades down and really creates a level of ownership in the business unit for using analytics to enable business success.”
Caterpillar serves industry segments including energy, transportation, construction, and mining, and has a unit for customer and dealer support. Ensuring that each group has its own analytics strategy enables Vawter’s organization to tailor analytics projects to maximize value for each unit — and to fit its analytics maturity level.
We want to make sure that we’re helping them to understand their data at the foundation and then advance them up the maturity curve.”
“
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
10
Success With Customer Data Depends on Keeping Customers’ Trust
Practitioners in our recent Data & Analytics survey have
gotten the memo on data security: Analytics depend
on data — and if that data is lost or stolen, or if cus-
tomers and partners become reluctant to share data because
of concerns over how it will be handled, the data-driven enter-
prise is at risk.
Common-sense security practices are widespread among sur-
vey respondents. A strong majority — 63% — either have or
are implementing a response plan in case their organization
suffers a data breach. Seven out of 10 either track or are cre-
ating the means to track where sensitive data is stored. Sim-
ilar majorities keep or are planning to create updated lists of
sensitive data they collect and train all employees in IT security
practices (see Figure 6).
The Pivotal Roles of Data Security and Privacy
Even more sophisticated measures are in place or underway
at most organizations. A slight majority have implemented or
are planning to deploy advanced analytics to predict cyber-
intrusion risks. More than half (54%) are using or implementing
a cybersecurity framework. And 37% have a chief information
security officer to lead these efforts, with another 18% plan-
ning or implementing that role (see Figure 7).
Analytics practitioners need to be aware of the environment
in which they do business. Even companies in the busi-
ness-to-business market that are not tied to negative publicity
— think of consumer data breaches or controversies about
the way social media networks manage user data — must note
customers’ heightened concerns about data use. “When
19%
Currently do this
Have responseplan in place
in case of data breach
Implementing Considering
44%
12%14%
Track where all data is stored
Keep updated list of sensitive
data types collected
Train all employees in IT
security risks and practices
Planning No activity
11%
47%
23%
11%10%10%
43%
20%
13% 13%11%
Currently do this
Apply advanced analytics to
predict cyber-intrusion risks
Implementing Considering
22%
16%15%
Use cybersecu-rity frameworks (e.g., PCI, NIST)
Employ a chief information
security o�cer
Planning No activity
33%
39%
15%
11%
37%
7%
13%
44%
20%
12%13%
11%
14%16%
20%
11%
33%
Figure 6: Data Breach Defenses Are Up
Figure 7: Security Frameworks and CISOs Take Hold
Organizations are moving toward solid, baseline data security practices, although many have yet to fully implement these measures.
Percentages may not equal 100 due to rounding.
A slight majority of survey respondents are increasing their data security maturity via implementation of security frameworks, and nearly half have or are hiring a CISO. A minority are using more sophisticated measures such as applying analytics and AI to security.
Percentages may not equal 100 due to rounding.
19%
Currently do this
Have responseplan in place
in case of data breach
Implementing Considering
44%
12%14%
Track where all data is stored
Keep updated list of sensitive
data types collected
Train all employees in IT
security risks and practices
Planning No activity
11%
47%
23%
11%10%10%
43%
20%
13% 13%11%
Currently do this
Apply advanced analytics to
predict cyber-intrusion risks
Implementing Considering
22%
16%15%
Use cybersecu-rity frameworks (e.g., PCI, NIST)
Employ a chief information
security o�cer
Planning No activity
33%
39%
15%
11%
37%
7%
13%
44%
20%
12%13%
11%
14%16%
20%
11%
33%
Currently do this Implementing Planning Considering No activity
Currently do this Implementing Planning Considering No activity
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
11
consumers see bad news stories out there about things that
happen, that can have a negative, spiraling effect on other
companies and other industries,” says Morgan Vawter, chief
analytics director at Caterpillar.
“We can’t do any of this work without trust with our customers
and the businesses that we serve,” Vawter adds. “And so we’re
making sure that we have clearly established rights to use the
information and ensuring that that lineage and that right and
that permission is very clear with every type of data that we
use and every data source that we access.”
While organizations are pursuing data security with some
urgency, their efforts on privacy lag. Just 41% of respondents
say they keep customers informed about data collection and
have internal controls in place. (see Figure 8).
The Opportunity to Build Customer Trust Based on Data
While some may view privacy regulations such as the Europe-
an Union General Data Protection Regulation (GDPR) as added
bureaucratic obligations, others see them as an opportunity to
build trust with customers. Susan Etlinger, an industry analyst
at the Altimeter Group, falls into the latter camp.
Etlinger observes that both business leaders and consumers
are becoming more aware of the many ways their data is used
and beginning to understand the impact that data use has on
their lives, their businesses, and society. As such, the widely
publicized 2018 compliance deadline for the GDPR caught
the attention of most organizations, with more than half
saying at the time of the survey that they had either finished
with compliance, were working on it, or were planning to work
on it (see Figure 9).
“This is actually a tremendous moment to reimagine what
data-centric organizations should look like and what the cus-
tomer experience should be,” she says. While organizations
must act in accordance with their fiduciary responsibilities
to comply with regulations, business leaders can also start to
imagine a world in which respecting privacy and building trust
works to their advantage.
In this environment, companies that figure out ways to use
data that provide value and enhance the trust of their custom-
ers will gain an edge. Given the option “to bank with a company
where you don’t know where your data’s going and you don’t
know how it’s being used or [to do business] with one who says,
‘You know, we actually want a partnership with you and we
want to be able to offer you the best possible experience, and
to do that, here’s what we ask and this is your choice,’” people
will be more inclined to choose the second, Etlinger says.
Trust Is Fragile — Handle With Care
Among leading analytics practitioners, there are a number
who strive to create such trusting relationships. As might be
expected, many can be found in industries like financial
41% Notify customers how we collect, use, and share their information, and have internal controls over how employees use the data
20%
25%
14%
Notify customers how we collect, use, and share their information
Implemented data privacy measures but have not yet communicated them externally
It’s not an issue we are concerned with
16%
27%
14%
25%
18%
We are fully compliant with GDPR
We are actively working on GDPR compliance
We are planning to comply with GDPR
GDPR is not a requirement but may guide our privacy policy
We have no plans for compliance
Privacy efforts lag security efforts, with just 41% keeping customers informed about data collection and use practices and also having internal controls in place.
Figure 8: Privacy Controls Have Room to Grow
Figure 9: GDPR Has Gained Attention
41% Notify customers how we collect, use, and share their information, and have internal controls over how employees use the data
20%
25%
14%
Notify customers how we collect, use, and share their information
Implemented data privacy measures but have not yet communicated them externally
It’s not an issue we are concerned with
16%
27%
14%
25%
18%
We are fully compliant with GDPR
We are actively working on GDPR compliance
We are planning to comply with GDPR
GDPR is not a requirement but may guide our privacy policy
We have no plans for compliance
GDPR has the attention of most survey respondents, with more than half saying they have finished or are planning or working on compliance.
Notify customers how we collect, use, and share their information, and have internal controls over how employees use the data
Notify customers how we collect, use, and share their information
Implemented data privacy measures but have not yet communicated them externally
It’s not an issue we are concerned with
We are fully compliant with GDPR
We are actively working on GDPR compliance
We are planning to comply with GDPR
GDPR is not a require-ment but may guide our privacy policy
We have no plans for compliance
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
12
services and health care. While these are, of course, governed
by stiffer regulations, success in these sectors has long
depended on earning and retaining deep trust, whether with a
customer’s life savings or a patient’s health.
At Sun Life Financial, data management leaders view GDPR as
a reinforcement of their existing practices, says Eric Monteiro,
senior vice president of client solutions. The financial services
company has a chief privacy officer, data breach notification
protocols, and a privacy impact assessment — all elements
required under GDPR. The company also has embarked on a
“plain and simple language initiative” that in the past year has
reviewed 500 standard letters customers receive with an eye
toward improving their clarity and minimizing legalese. The
prime reason for these activities — maintaining customer con-
fidence — predated the new rules.
“People give us their grandkids’ and great-grandkids’ money.
So they really need to be able to trust us,” Monteiro says. “Trust
is one of those things that is very hard to build and very easy to
lose, as we have all seen in recent events in the media.”
Monteiro says Sun Life follows a policy of providing value for
whatever data clients share. The company’s digital benefits as-
sistant, called Ella, is an example. Ella uses a set of predictive
models to “nudge” clients to take actions that are in their best
interest, such as maximizing retirement contributions (if fi-
nancial data shows the client is not) and using ancillary health
care services (like a chiropractor or wellness program) that can
improve the person’s quality of life based on, for example, the
stress inherent in the person’s job. The firm’s satisfaction scores
increase significantly — 27 points in a Net Promoter Score scale
— when Ella engages proactively with these clients, he says.
Health care is an industry in which leaders have experience in
guarding customer data. Dr. Timothy Crone, medical director,
business intelligence and enterprise analytics, at the Cleveland
Clinic, says the organization takes a cautious approach with
its data management, even as opportunities to share data with
outside parties promise to yield new health insights that could
benefit patients — and even as some enthusiastic patients want
to share more of their data.
Why? Because, as Crone explains, even if patients trust health
care providers more than some other businesses that make
headlines with data breaches, “I think that, at this point, that’s
a privilege we still have the opportunity to lose.”
That is a prudent approach. What the new GDPR requirements
are effectively doing is pushing retailers and other companies
to be more like organizations in health care and financial ser-
vices, says Dean Abbott, co-founder and chief data scientist at
SmarterHQ.
“GDPR does very good things for consumers, of course, but
there’s a lot of best practices in the regulations,” Abbott says,
noting that the additional transparency represents a strong
step. It means that data scientists have to be ready to not only
justify their use of customer data in algorithms, but be ready
to communicate how and why they are doing it. The impor-
tance of that transparency and data stewardship only grows as
organizations’ use of analytics evolves to AI by automating the
models and adding learning elements.
And that means analytics professionals need to monitor the
evolving social contract for using data to customers’ benefit.
“We view ourselves as a customer-first company, and we are
nothing without our customers’ success — and then their trust,
ultimately,” says Caterpillar’s Vawter. “And if we’re not creating
value from the data that we’re collecting from them, then we
shouldn’t be using it at all.” l
“We view ourselves as a customer- first company, and we are nothing without our customers’ success — and then their trust, ultimately.” MORGAN VAWTER, CATERPILLAR
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
13
DataSF Teaches the Art of Asking Analytical Questions
I N D U S T R Y S N A P S H O T
DataSF, the analytics group for the city and county of San Francisco, is teaching people throughout the government to find opportunities for advanced analytics. It’s doing that by teaching local officials how to frame questions that tap data and the answers that can improve public services.
That approach has led to projects such as work with the Department of Public Health to understand what is driving the cost of its mental health programs. DataSF also helped the city administrator’s office find opportunities for fleet vehicle sharing to improve use of vehicles, leading to both lower costs for the city and greener management of the fleet.
“It’s about developing the organizational muscle across all those different service lines to ask questions that are amenable to data science,” explains Joy Bonaguro, who was chief data officer for the city and county of San Francisco at the time she was interviewed for this report. The objective is to help clients within government spot opportunities to use AI and data science within their own verticals.
“We’ve developed something called a project typology that we use to help solicit and define data science questions with our departments. So we don’t say, ‘Hey, do you want to do AI and data science?’ We say, ‘Here are the kinds of questions, and here are a bunch of examples that we can help you answer,’” Bonaguro explains. “If we train our departments to spot data science opportunities, then that’s how we spread it throughout the organization.”
Bonaguro credits the work of others in the public sector, including New Orleans’ NOLA-lytics, which developed an analytics topology that speaks to common business problems. DataSF has posted an online information sheet that categorizes the kinds of problems data science can solve, such as finding the needle in a haystack (figuring out where to direct limited resources), prioritizing a backlog, flagging important items early for action, A/B testing (to find out which communication style works best), and optimizing resources.
GO
VE
RN
ME
NT
Joy Bonaguro, former chief data officer,
city and county of San Francisco
If we train our departments to spot data science opportunities, then that’s how we spread it throughout the organization.”
“
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
14
Cross-Functional Teamwork Improves Predictive Models at Barclays US
I N D U S T R Y S N A P S H O T
Putting analytics experts shoulder to shoulder with business domain experts doesn’t just build a stronger culture. At Barclays US, it has built better analytics.
The primary focus of analytics at the bank and credit card issuer is the customer journey — getting to the right customer at the right time with the right kind of offer, according to Vishal Morde, vice president of data science and advanced analytics. “If you can actually achieve that, it will make your business more profitable. And your customers will be happy because they’re getting what they’re really looking for.”
A key component of the company’s analytics approach is to tap business domain expertise throughout a project, Morde says.
Barclays US has integrated business people into the analytics process by creating a data sci-ence lab and setting up a cross-functional team that includes business owners. “They were part of the creation of this data science lab,” Morde explains. “Because they’re fairly inte-grated upfront, we could actually now set up this whole environment, set up projects which would actually deliver value and help them to solve the business problem.”
For example, a project to create a better predictive model for determining who will respond to a certain kind of offer included both data scientists and acquisition marketing staff. “The data scientist would say, ‘Hey, I’m looking at this data, and there’s some seasonality to it.’ And a business person will say, ‘Yeah, that makes sense. It has to be the holiday season. That’s where people start responding.’”
Because interactions like this were happening earlier in the process, before data scientists actually produced a model, the teams were able to produce better models more quickly, Morde says. By tapping the consumer insight and some of the anecdotal hypotheses that business experts have, and testing them out with advanced data science methods, Barclays US was able to improve prediction power significantly — in some cases by 50%, he adds.
Wins like that don’t come from just data, methods, or analytics tools, Morde says: “It was that we were able to transfer some of the domain knowledge from the marketing folks into our models. You’re actually incorporating years and years of expert knowledge that people gathered about consumer behavior and consumer needs and wants.”
FIN
AN
CIA
LS
ER
VIC
ES
Vishal Morde, vice president of data
science and advanced analytics, Barclays US
“You’re actually incorporating years and years of expert knowledge that people gathered about consumer behavior and consumer needs and wants.”
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
Building Trust in Innovation by Creating a Culture of Inquiry and Experimentation
Game-changing insights can result from investments in
data infrastructure, talent, and technology. But simply
generating analytical findings is not the ultimate goal
of the analytics leaders we interviewed for this year’s Data & An-
alytics report. All those investments truly pay off when people
throughout an organization accept those insights and put them
to work to make decisions, redesign business processes, and re-
think strategy. So, what are some tactics used to create a culture
in which people ask questions, seek data and analytics that can
help answer those questions, and then apply the results?
Barclays US appointed an analytics leader for each business unit
and installed cross-functional teams made up of analytics and
business experts in a data science lab. Caterpillar runs an annual
“Analytics Now” conference for stakeholders across the compa-
ny to learn about capabilities and showcase successful projects,
while seeking input and collaboration from an advisory council
of engineers, talent managers, and other domain experts. The
Cleveland Clinic has set up its own council, with an open door to
participate and provide feedback on its analytics program.
Our survey investigated a range of practices associated with
building a culture of analytics, such as actions and messaging
from leadership, workforce data literacy efforts, and organiza-
tional choices. On the whole, those who report the most activity
on these fronts are also more likely to exhibit the most trust
in data and analytics and be on the winning side of the “utility
gap,” meaning they not only have access to data but have the
right data to inform their business decisions.
Two organizational factors emerged that may have a bearing
on driving analytics maturity:
1. One is having a clear leader for the company’s analytics
effort: 33% of those who report that their organization has a chief
data officer or a chief analytics officer are more likely to say they
frequently or always have the data they need for decisions. The
same correlation is present for the 39% of respondents who report
that their organizations have centralized data analytics functions.
(Despite this finding, one size may not fit all: Some analytics
leaders find a distributed approach is right for their enterprise.)
2. The survey also found that in many organizations, leaders
are playing an important role in using and championing analyt-
ics. However, leaders’ actions may be slightly louder than their
words: They appear to be more likely to seek out data and use it
in decision-making than to champion analytics or credit it with
delivering business results (see Figure 10). Leadership support is
also softer when it comes to prioritizing investment in analytics.
Always Often Sometimes Rarely Never
11%25%
33%21%
9%
13%29%
30%20%
8%
19%32%
30%15%
4%
17%36%
33%11%
3%
19%40%
29%
3%9%
3%9%
16%40%
32%
Prioritize investments in analytics tools
Credit positive business outcomes
to analytics in internal messages
or presentations
Champion the value and use of analytics
Incorporate data and analytics in
decision-making
Seek data and analytics support
decisions
Understand insights presented
by analytics specialists
Figure 10: Leaders Set the Tone for Analytics Adoption
How often do leaders:
Leaders at the majority of companies often champion the value of data and actively seek to apply it when making decisions.
Percentages may not equal 100 due to rounding.
Always Often Sometimes Rarely Never
Prioritize investments in analytics tools
Credit positive business outcomes
to analytics in internal messages
or presentations
Champion the value and use of
analytics
Incorporate data and analytics in
decision-making
Seek data and analytics
to support decisions
Understand insights
presented by analytics
specialists
15
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
Driving Data Literacy Through the Workforce
One area where leadership might do more is analytics skills in
the workforce: This was cited as a top challenge to innovating
by 39% of respondents, second only to competition from other
priorities. We found that while a range of efforts to build data
literacy in the workforce is at least in the planning stages at a
majority of responding organizations, fewer than one in five are
actively engaged in that activity (see Figure 11). Slightly more
are helping analytics experts build business domain expertise.
Encouraging a data-driven culture means encouraging people
to use the data the business is collecting instead of siloing
it away in the IT department, says Kirk Borne, principal data
scientist and executive advisor at consultancy Booz Allen Ham-
ilton. “This culture of experimentation supports an innovation
culture in a business. It’s where people have the freedom to
ask, ‘Given our data, how can we do better?’” he says.
One example of bringing data literacy to the front lines of busi-
ness comes from Jeanne Ross, a principal research scientist at
the MIT Center for Information Systems Research.
The center studied the 7-Eleven convenience store chain in
Japan, which employed counselors to help 200,000 salespeo-
ple interpret and learn from the data the company collected
about daily sales. Sales staff know a key success factor is their
shop’s ability to have rapid inventory turnover, so they monitor
sales of different items in their assigned section of the store.
Counselors visited the stores to teach the salespeople how to
think analytically, using data about sales of different items in
recent days compared to the year earlier and compared to days
with similar weather patterns.
“The counselors say to them, ‘So, what did you hypothesize
about what you’d sell last week?’ They know the answer to the
question because it’s what they ordered,” Ross says. “Then they
say, ‘How did you do? How did your hypotheses turn out?’ Well,
they have the answer to that question sitting in front of them,
too. They have the data.” Then the counselors and sales staff
discuss strategies for improving sales.
Fostering Collaboration Drives Culture Change
It’s when the analytics expert meets the business domain and
true collaboration ensues that the culture really begins to
transform, according to virtually all the practitioners we inter-
viewed for this report. Interestingly, our survey respondents
“Encouraging a data-driven culture means encouraging people to use the data the business is collecting instead of siloing it away in the IT department.” KIRK BORNE, BOOZ ALLEN HAMILTON
Currently do this Implementing
17%18%
19%18%
21%19%
15%16%
17%18%
30%19%
16%14%
17%19%
22%20%
15%
Line-of-business experts receive
training or immersion in
analytics
Analytics specialists receive
training or immersion in
operational areas
Training programs are widely available to develop data and
analytical skills
Workforce data literacy is regularly
assessed
Internal messaging promotes data
literacy as a valued skill
Planning
Considering No activity
30%
29%
17%
35%
17%25%
Many organizations have an opportunity to do more to tackle the analytics skills shortage. The good news is that a majority are taking action to build a data-driven workforce, with programs either running or in the planning or implementation stages.
Percentages may not equal 100 due to rounding.
Figure 11: Educate to Innovate
Currently do this Implementing Planning Considering No activity
Line-of-business experts receive
training or immersion in analytics
Analytics specialists receive training or immersion in
operational areas
Training programs are widely available to develop data and
analytical skills
Workforce data literacy is regularly
assessed
Internal messaging promotes data
literacy as a valued skill
16
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
17
by a slight margin point to individuals and teams in operating
units as being most likely to drive innovation with emerging
technologies, ahead of top leadership and even IT (see Figure
12). That underscores why building ties with the business may
be a winning strategy for analytics leaders.
Establishing and maintaining a culture that embraces analytics
starts with identifying opportunities and use cases that will
make a difference.
At Airbnb, a platform that has disrupted the hospitality indus-
try, everything begins with analytics and a strong top-down
data culture, according to Yash Kandyala, head of global busi-
ness analytics for community support.
Nonetheless, Kandyala still needs to have the right data and en-
gage the right people in order to find opportunities for analyt-
ics to drive strategy. At Airbnb, it’s a multidirectional dynamic.
Sometimes colleagues come with questions and want answers.
At other times, his team may notice an important insight in the
data and seek to gain executives’ attention and support.
Kandyala says gaining influence for analytics insights gen-
erated by his team can be challenging: “You may have a very
cool idea, but if it’s not tying back to key stakeholders’ goals or
strategy, then there’s going to be very little attention paid to
it.” He seeks to identify projects that can help drive important
goals for the company. “That can go a long way to create those
opportunities and projects which will be based on analytics as
the backbone and that are tied to a business outcome.”
Analytics Expertise: Centralize vs. Decentralize
As noted, centralization isn’t for everyone. At Sun Life Financial,
embedding analytics capabilities in the business rather than
employing a centralized analytics function ensures strategic
alignment and transparency.
“The idea of centralizing and creating a big function and putting
a lot of money in it centrally is very alluring, because it sounds
like you are going to solve all of your data problems,” says Eric
Monteiro, senior vice president of client solutions. “The benefit
of not having done that is that it’s pretty visible to us what the
impact is when the analytics is with the business and for the
business, and linked into the processes that are required.”
Sun Life analytics and business leaders have codified their col-
laboration on a case-by-case basis. Every analytics program
requires developing an ROI case and winning budget approval from
business decision-makers before it starts. There are also regular
progress reports with business owners to check on results.
Despite the decentralized effort, Monteiro says results can
grow from one project into a major operational area. That’s how
the company’s digital benefits assistant started. It was origi-
nally an analytics experiment to personalize recommendations
for customers, such as tips on retirement contributions. It’s
been so useful that it’s become a major undertaking. “We’ve got
dozens of people that work on it, and it’s a digital channel and
reports like any distribution channel,” he says.
Figure 12: Innovation Is Often a Grassroots Effort
Individuals close to specific business needs are often the drivers of innovation around emerging technologies.
Who is most likely to champion the use of emerging technologies such as AI/machine learning, internet of things (IoT), and blockchain?
24%Top leadership
26%
9%
13%
23%
5%
Individual/teamsin operating units
Marketing
R&D
IT
Other
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
18
There is a challenge associated with the lack of a centralized
analytics function, Monteiro adds: what to do when there is a
major investment required that goes beyond a typical project.
For example, Sun Life is looking to rearchitect some of its data
so it can move to the cloud, which requires more investment in
architecture and infrastructure. “The incrementalism is a risk
that we have to manage,” he says.
Communication and Education Encourage an Analytics Mindset
Advancing a culture of innovation requires not just being heard
but listening. Joy Bonaguro, who was chief data officer for the
city and county of San Francisco at the time she was inter-
viewed for this report, says much of the work of data science
is about culture and change management, helping internal cli-
ents understand the value of data and how it can lead to new
and innovative ways of thinking. A
key element of the effort is making
sure that data scientists listen.
“The way we’ve approached cul-
ture change, driving data use in the
city, has been to really spend time
to understand the challenges and
barriers to using data across the
city. Then, let’s shape our services
around that so people feel heard,”
Bonaguro says. “And when people
feel heard, then they start to trust
you and they want to work with you
on new things, including things that
are maybe a little riskier or scarier,
like data science projects.”
Getting to that point came after a
number of efforts, including pro-
viding training via the organiza-
tion’s Data Academy (see Figure 13),
demonstration projects like dash-
boards, and the launch of an open
data portal, Bonaguro adds. Over
time, through these educational efforts and by having data
experts working side by side with staff, the agency’s internal
clients have gained the tools to identify the appropriate data
science projects for them and then apply for data science
resources. This approach also helps analytics professionals,
who serve such a diversity of internal customers that they
can’t also be domain experts — Bonaguro points out that San
Francisco’s departments run from “A to Z, an airport to a zoo.”
One success was a DataSF project with the county Department
of Public Health to investigate why participation rates had
fallen by 16% among mothers and babies in the federal Wom-
en, Infants, and Children (WIC) nutrition program from 2011
to 2017. After analyzing six years of data about participation,
program staff interactions with mothers, payments issued, and
Strongly Agree
DataSF's Data Academy Assessment: Leaders Set the Tone for Analytics Adoption
Agree Neither Agree nor Disagree Disagree Strongly Disagree
32% 48% 15%
0% 20% 40% 60% 80% 100%
18% 28% 23%
0% 20% 40% 60% 80% 100%
Daily Weekly Monthly Rarely Never
22% 9%
Figure 13: DataSF’s Data Academy Assessment
Do you feel that your skills improved after taking this Data Academy course?
How often do you use the information or skills you learned in your own work?
DataSF, the analytics group for the city and county of San Francisco, is expanding data literacy throughout the agencies it serves via its Data Academy. It shares success metrics — such as attendees’ assessments of skills gained and how frequently those are applied — via a public dashboard.
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
19
demographic details, the team developed a model that showed
certain families were more likely to drop out. The results
prompted San Francisco’s WIC program to change its outreach
efforts to encourage more participation.
Such collaborations lead to high-quality results and real con-
nections with internal customers because both parties are in-
vesting in the outcome. “This is where the model of the data
scientist getting the data and disappearing off into a cave and
resulting in some magical, perfect model is a terrible mindset,”
she says. “We set the expectations with our clients that this is
going to take time and you’re really going to have to engage.”
And it also serves to educate people who are not data experts
about the work of analytics and its potential impact — and builds
trust in data scientists’ work as they collaborate to find insights.
Bonaguro’s data science shop serves San Francisco’s range of
“airport to zoo” departments with a tight team of just five full-
time people, with at most three working on the data science
service. But the same collaborative principles apply at larger en-
terprises like Barclays US, Caterpillar, and the Cleveland Clinic.
Vishal Morde, vice president of data science and advanced
analytics at Barclays US, says the company’s data science lab
includes business experts who work with analytics experts to
set up projects to solve business problems. “Both sides need to
come together; both sides need to actually make sure that they
understand each other’s perspective,” Morde says. “They un-
derstand each of the challenges and come up with a more kind
of unified approach, rather than working in silos and saying
that, ‘Oh, somebody else needs to do this job for me.’”
At Caterpillar, culture-building takes the form of ongoing
training sessions like “How to Be an Analytics Champion in the
Business” as well as its annual analytics conference, says Mor-
gan Vawter, chief analytics director. “We certainly understand
not everybody wants to be an analytics professional, nor do we
need them to be. But we do want to help them to understand
the power of the data that they already have access to, and so
we provide a lot of training around that,” Vawter says.
The Cleveland Clinic takes a similar tack. In addition to mak-
ing its advisory council open to all, its analytics organization is
identifying opportunities to collaborate with different groups
throughout the enterprise, says Chris Donovan, executive di-
rector of enterprise information management and analytics.
“We’re very focused on delivering value and specific projects,
and making efforts that we have key sponsors for across the
organization. So, as we lay out our road map, we try to make
sure that we’re working with our cardiovascular institute in
this space. We’re working with our cancer institute in this
space, and our strategy team and our marketing team,” Dono-
van says. Showing each group what his team can do builds sup-
port: “That really engages those folks to be champions for the
program and creates this virtuous cycle of where they recog-
nize the challenges around data, and understanding the need
for governance. And it creates a nice feedback loop for us.”
To a veteran of data science discussions like Eric Siegel, the
efforts and experiences of companies like these demonstrate
how far the field has come. Siegel, author of Predictive Analyt-
ics: The Power to Predict Who Will Click, Buy, Lie, or Die, says
technical experts have always faced challenges in communi-
cating the value of their work to executives, winning approv-
al for projects, and convincing their colleagues to trust and
actually use the results generated by core analytics. What’s
changed are the types of champions he sees rising nowadays to
speak about their work at industry conferences he organizes,
such as Predictive Analytics World and Deep Learning World.
Increasingly, C-suite executives, vice presidents, and directors
of business functions are contributing their voices and exper-
tise to events in the analytics arena, Siegel says. In other words,
leaders whose job titles don’t have technology or analytics in
them are making the case for analytics — and building support
for a data-driven culture. l
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
20
About the Research
MIT SMR Connections conducted an online survey of MIT Sloan Management Review subscribers in May 2018, drawing 2,413
respondents. Respondents represent a broad range of functions and industries, with more than 70% identifying themselves as
being in management, C-suite, or board roles. The survey included respondents from all regions of the world.
To provide a rich context for discussion of the quantitative research results, we interviewed analytics experts including practi-
tioners, consultants, and academics. These individuals provided insight into good practices and examples of the steps leading
organizations are taking to advance analytical maturity and build data-driven cultures.
Dean Abbott, co-founder and chief data scientist, SmarterHQ
Joy Bonaguro, former chief data officer, city and county of San Francisco
Kirk Borne, principal data scientist and executive adviser, Booz Allen Hamilton
Timothy Crone, MD, medical director, business intelligence and enterprise analytics, Cleveland Clinic
Chris Donovan, executive director of enterprise information management and analytics, Cleveland Clinic
Susan Etlinger, industry analyst, Altimeter Group
Caroline Viola Fry, PhD candidate, MIT Sloan School of Management
Michael S. Goldberg, independent writer and researcher
Christina Hoy, vice president, corporate business information and analytics, Workplace Safety and Insurance Board Ontario
Yash Kandyala, head of global business analytics for community support, Airbnb
David Loshin, principal consultant, Knowledge Integrity
Eric Monteiro, senior vice president of client solutions, Sun Life Financial Canada
Vishal Morde, vice president of data science and advanced analytics, Barclays US
Jeanne Ross, principal research scientist, MIT Sloan Center for Information Systems Research
Beatriz Sanz Saiz, global data and analytics leader for Advisory Services, EY
Eric Siegel, author, Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die, and founder, Predictive Analytics World conference
Morgan Vawter, chief analytics director, Caterpillar
Acknowledgments
CUSTOM RESEARCH REPORT — DATA, ANALYTICS, AND AI: HOW TRUST DELIVERS VALUE
21
The annual MIT Sloan Management Review Custom Data & Analytics survey highlights how the most analytically innovative organizations are implementing new tech-
nologies — most recently, AI-based techniques like machine learning and natural language processing.
This latest report focuses on an aspect of data and analytics that hits home for SAS, not just from a technical standpoint, but from an emotional one: trust.
To rely on analytics for decision-making, you have to trust your data. Successful organizations also must trust in their ability to innovate and create a culture where asking questions is welcomed and encouraged.
Then there is the external trust factor — highlighted with the European Union’s recent implementation of the General Data Protection Regulation — in which security and privacy con-cerns are critical for gaining and keeping customers’ faith that you’ll do right by their data.
So why does that connect with us here at SAS? Our mission is “to empower and inspire with the most trusted analytics.” And for more than 40 years, our purpose has been to use curiosity as the driving force behind progress.
Not surprisingly, organizations that have advanced their analytics practice to apply machine learning, AI, and automated analytics to workflows have a similar mindset.
They are more likely to take actions to build data quality, data security, and a culture of data-driven innovation. With reliable data, these analytically mature organizations can move beyond basic BI and dashboards.
The “robust majority” are getting their security measures in order to foster external trust. Privacy efforts, while lagging slightly behind security, are also a priority — obviously driven in part by GDPR compliance.
Eventually, this is another place where data management and AI will meet, with organizations trusting their data and analytics enough to make them part of the foundation for protecting other data.
But none of this happens unless company culture supports it, which could mean anything from hiring a chief data officer or chief analytics officer and having other leaders champion analytics to building skills and strategy with existing talent.
All in all, the results of the survey demonstrate a truth we’ve always believed in: Technology and trust go hand in hand, especially when it comes to the future of data.
Why Advancing Technology Demands Building Trust
Randy Guard, executive vice president and chief marketing officer, SAS
S P O N S O R ’ S V I E W P O I N T
About SASThrough innovative analytics, business intelligence, and data management software and services, SAS helps customers make better decisions faster. Since 1976, SAS has been giving customers around the world THE POWER TO KNOW®.
To learn more about how technology and trust go hand in hand, visit us at www.SAS.com/innovation.